一种多信息作用下的激光回环检测算法

A Lidar Loop Detection Algorithm with Multiple Information Effects

为提高回环检测的精度,同时满足实时性的需求,提出了一种多信息作用下的激光回环检测算法。首先,利用惯性测量单元(inertial measurement unit,IMU)信息得到每个点云帧之间的相对运动;然后,提取点云中的高度信息、强度信息以及数量信息并归一化处理;在上下文扫描(scan context,SC)描述符的基础上,定义了融合IMU与点云信息的上下文扫描(IMU and point scan context,IPSC)描述符以及对应的相似度函数;最后,引入改进的正态分布变换(normal distributions transform,NDT)算法进行综合相似度判断。针对所提出的算法在多个数据集上进行实验,结果表明,该算法在回环检测时具有较高的精度,能够在不同场景下进行识别,同时满足了系统的实时性要求,为激光回环检测提供了一种新的方法。

To improve the accuracy of loop detection and meet the real-time requirements,a lidar loop detection algorithm with multiple information is proposed.Firstly,the relative motion between each point cloud are obtained by using the inertial measurement unit(IMU)information.Then,the height,intensity and quantity information of the point cloud are extracted and normalised;based on the Scan Context(SC)descriptor,IMU and Point Scan Context(IPSC)descriptors are defined and the corresponding similarity functions are defined.Finally,an improved Normal Distributions Transform(NDT)algorithm is introduced for comprehensive similarity determination.The proposed method has been experimented on several data sets,and the results show that the algorithm has high accuracy in loop detection,and can identify in different scenarios,and can also meet the real-time requirements of the system,providing a new method for lidar loop detection.